WP/15/89

Fair Weather or Foul? The Macroeconomic Effects of El Nio

by Paul Cashin, Kamiar Mohaddes and Mehdi Raissi

2015 International Monetary Fund WP/15/89

IMF Working Paper

Asia and Pacific Department

Fair Weather or Foul? The Macroeconomic Effects of El Nio*1

Prepared by Paul Cashin, Kamiar Mohaddes** and Mehdi Raissi

Authorized for distribution by Paul Cashin

April 2015

Abstract

This paper employs a dynamic multi-country framework to analyze the international macroeconomic transmission of El Nio weather shocks. This framework comprises 21 country/region-specific models, estimated over the period 1979Q2 to 2013Q1, and accounts for not only direct exposures of countries to El Nio shocks but also indirect effects through third-markets. We contribute to the climate-macroeconomy literature by exploiting exogenous variation in El Nio weather events over time, and their impact on different regions cross-sectionally, to causatively identify the effects of El Nio shocks on growth, inflation, energy and non-fuel commodity prices. The results show that there are considerable heterogeneities in the responses of different countries to El Nio shocks. While Australia, Chile, Indonesia, India, Japan, New Zealand and South Africa face a short-lived fall in economic activity in response to an El Nio shock, for other countries (including the United States and European region), an El Nio occurrence has a growth-enhancing effect. Furthermore, most countries in our sample experience short-run inflationary pressures as both energy and non-fuel commodity prices increase. Given these findings, macroeconomic policy formulation should take into consideration the likelihood and effects of El Nio weather episodes.

JEL Classification Numbers: C32, F44, O13, Q54.

Keywords: El Nio weather shocks, oil and non-fuel commodity prices, global macroeconometric modeling, international business cycle.

Authors E-Mail Addresses: pcashin@imf.org; km418@cam.ac.uk; mraissi@imf.org.

* We are grateful to Tiago Cavalcanti, Hamid Davoodi, Govil Manoj, Rakesh Mohan, Sam Ouliaris, HashemPesaran, Vicki Plater, Ajay Shah, Ian South, Garima Vasishtha, Rick van der Ploeg, Yuan Yepez and seminar participants at the IMF's Asia and Pacific Department Discussion Forum, the IMF's Institute for Capacity Development, Oxford University, and the 13th Research Meeting of the National Institute of Public Finance and Policy and the Department of Economic Affairs of the Ministry of Finance in India for constructive comments and suggestions. The views expressed in this paper are those of the authors and do not necessarily represent those of the International Monetary Fund or IMF policy. ** Faculty of Economics and Girton College, University of Cambridge, United Kingdom.

This Working Paper should not be reported as representing the views of the IMF.

The views expressed in this Working Paper are those of the author(s) and do not necessarily represent those of the IMF or IMF policy. Working Papers describe research in progress by the author(s) and are published to elicit comments and to further debate.

2

Contents Page

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

II. The Southern Oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

III. Modelling the Climate-Macroeconomy Relationship in a Global Context . . . . . . . . . 8 A. The Global VAR (GVAR) Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 B. Model Specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

IV. Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 A. The Effects of El Nio on Real Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16 B. The Effects of El Nio on Real Commodity Prices. .. . . . . . . . . . . . . . . . . . . 21 C. The Effects of El Nio on Inflation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 D. Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

V. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Tables 1. Countries and Regions in the GVAR Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2. Lag Orders of the Country-Specific VARX*(p,q) Models Together with the Number

of Cointegrating Relations (r) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15 3. The Effects of an El Nio Shock on Real GDP Growth (in percent) . . . . . . . . . . . . .17 4. Share of Primary Sector in GDP (in percent), Averages over 2004-2013 . . . . . . . . . 17 5. Trade Weights, Averages over 20092011 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 6. The Effects of an El Nio Shock on Real Commodity Prices (in percent) . . . . . . . . .22 7. The Effects of an El Nio Shock on Inflation (in percent). . . . . . . . . . . . . . . . . . . . . 24

Figures 1. Growth and Inflation Following the 1997/98 El Nio Episode . . . . . . . . . . . . . . . . . . 4 2. Southern Oscillation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3. Global Climatological Effects of El Nino . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 4. Southern Oscillation Index (Anomalies), 1979M42014M12 . . . . . . . . . . . . . . . . . . .7 5. Food Weight in CPI Basket and Inflation Responses . . . . . . . . . . . . . . . . . . . . . . . . .24

3I. INTRODUCTION

A rapidly growing literature investigates the relationship between climate (temperature,precipitation, storms, and other aspects of the weather) and economic performance(agricultural production, labor productivity, commodity prices, health, conflict, and economicgrowth)see the recent surveys by Dell et al. (2014) and Tol (2009). This is important as acareful understanding of the climate-economy relationship is essential to the effective designof appropriate institutions and macroeconomic policies, as well as enabling forecasts of howfuture changes in climate will affect economic activity. However, a key challenge in studyingsuch a relationship is "identification", i.e. distinguishing the effects of climate on economicactivity from many other characteristics potentially covarying with it. We contribute to theclimate-economy literature by exploiting the exogenous variation in weather-related events(with a special focus on El Nio1) over time, and their impact on different regionscross-sectionally, to causatively identify the effects of El Nio weather shocks on growth,inflation, energy and non-fuel commodity prices within a compact model of the globaleconomy.

Our focus on El Nio weather events is motivated by growing concerns about their effects notonly on the global climate system, but also on commodity prices and the macroeconomy ofdifferent countriessee Figure 1 for the evolution of growth and inflation across countriesfollowing the most recent strong El Nio episode which started in 1997Q2 and ended in1998Q1. These extreme weather conditions can constrain the supply of rain-drivenagricultural commodities, create food-price and generalized inflation, and may trigger socialunrest in commodity-dependent countries that primarily rely on imported food. It has beensuggested, by both historians and economists, that El Nio shocks may even have played arole in a substantial number of civil conflicts, see Hsiang et al. (2011). To analyze themacroeconomic transmission of El Nio shocks, both nationally and internationally, weemploy a dynamic multi-country framework (combining time series, panel data, and factoranalysis techniques), which takes into account economic interlinkages and spillovers that existbetween different regions. It also controls for macroeconomic determinants of energy andnon-fuel commodity prices, thereby disentangling the El Nio shock from many otherpossible sources of omitted variable bias. This is crucial, given the global dimension ofcommodity-price dynamics, and the interrelated macroeconomic performance of mostcountries.

1El Nio is a band of above-average ocean surface temperatures that periodically develops off the Pacific coastof South America, and causes major climatological changes around the world.

4Figure 1. Growth and Inflation Following the 1997/98 El Nio Episode

(a) Growth (b) Inflation

Source: Authors calculations.Notes: Year on year percent change in growth and inflation following the 1997/98 El Nio episode.

Despite their importance, the macroeconomic effects of the most recent strong El Nio eventsof 1982/83 and 1997/98, along with the more frequent occurrences of weak El Nios, areunder-studied. There are a number of papers looking at the effects of El Nio on: particularcountries, for example, Australia and the United States (Changnon 1999 and Debelle andStevens 1995); a particular sector, for instance, agriculture and mining (Adams et al. 1995 andSolow et al. 1998); or particular commodity markets, including coffee, corn, and soybean(Handler and Handler 1983, Iizumi et al. 2014, and Ubilava 2012). Regarding the economicimportance of El Nio events, Brunner (2002) argues that the Southern Oscillation (ENSO)cycle can explain about 1020 percent of the variation in the GDP growth and inflation of G-7economies, and about 20% of real commodity price movements over the period 19631997.2

He shows that a one-standard-deviation positive shock to ENSO raises real commodity priceinflation by about 3.5 to 4 percentage points (but this effect is only statistically significant inthe second quarter following such a shock), and although the median responses of G-7economies aggregate CPI inflation and GDP growth are positive in the first four quarters,they are both in fact statistically insignificant. While Brunner (2002) focuses on the economiceffects El Nio shocks over time (only taking advantage of the temporal dimension of thedata), his sample is mostly restricted to regions which are not directly affected by El Nio.

2The Southern Oscillation index (SOI) measures air-pressure differentials in the South Pacific (between Tahitiand Darwin). Deviations of the SOI index from their historical averages indicate the presence of El Nio (warmphase of the Southern Oscillation cycle) or La Nia (cold phase of the Southern Oscillation cycle) eventsseeSection II. for more details.

5We contribute to the literature that assesses the macroeconomic effects of weather shocks inseveral dimensions, including a novel multi-country methodology. Our modelling frameworkaccounts for the effects of common factors (whether observed or unobserved), and ensuresthat the El Nio-economy relationship is identified from idiosyncratic local characteristics(using both time-series and cross-section dimensions of the data). To the extent that El Nioevents are exogenously determined, reverse causation is unlikely to be a concern in ourempirical analysis. Nevertheless, we allow for a range of endogenous control regressors,where country-specific variables are affected by El Nio shocks and possibly simultaneouslydetermined by other observed or unobserved factors. We also have a different macroeconomicemphasiswhile Brunner (2002) mainly focuses on the effects of El Nio on commodityprices, we concentrate on the implications of El Nio for national economic growth andinflation, in addition to global energy and non-fuel commodity prices. Moreover, we study theeffects of El Nio shocks on 21 individual countries/regions (some of which are directlyaffected by El Nio) in an interlinked and compact model of the world economy, rather thanfocusing on an aggregate measure of global growth and inflation (which Brunner 2002 takesto be those of G-7 economies). Furthermore, we explicitly take into account the economicinterlinkages and spillovers that exist between different regions in our interconnectedframework (which may also shape the responses of different macroeconomic variables to ElNio shocks), rather than undertaking a country-by-country analysis. Finally, we contribute tothe Global VAR (GVAR) literature that mostly relies on reduced-form impulse-responseanalysis by introducing El Nio as a dominant and causal variable in our framework.

Our framework comprises 21 country/region-specific models, among which is a singleEuropean region. These individual-economy models are solved in a global setting where coremacroeconomic variables of each economy are related to corresponding foreign variables anda set of global factorsincluding a measure of El Nio intensity as a dominant unit. Themodel has the following variables: real GDP, inflation, real exchange rate, short-term andlong-term interest rates, real energy and non-fuel commodity prices, and the SouthernOscillation index (SOI) anomalies as a measure of the magnitude of El Nio. This frameworkaccounts for not only direct exposures of countries to El Nio shocks but also indirect effectsthrough third-markets; see Dees et al. (2007) and Pesaran et al. (2007). We estimate the 21individual VARX* models over the period 1979Q22013Q1. Having solved the Global VARmodel, we examine the effect of El Nio shocks on the macroeconomic variables of differentcountries (especially those that are most susceptible to this weather phenomenon).3

3The GVAR methodology is a novel approach to global macroeconomic modeling as it combines time series,panel data, and factor analysis techniques to address the curse of dimensionality problem in large models, and is

6Contrary to the findings of earlier studies, the results of our dynamic multi-country model ofthe world economy indicate that the economic consequences of El Nio shocks are large,statistically significant, and highly heterogeneous across different regions. While Australia,Chile, Indonesia, India, Japan, New Zealand and South Africa face a short-lived fall ineconomic activity in response to a typical El Nio shock, for other countries, an El Nio eventhas a growth-enhancing effect; some (for instance the United States) due to direct effectswhile others (for instance the European region) through positive spillovers from major tradingpartners.4 Overall, the larger the geographical area of a country, the smaller the primarysectors share in national GDP, and the more diversified the economy is, the smaller is theimpact of El Nio shocks on GDP growth. Furthermore, most countries in our sampleexperience short-run inflationary pressures following an El Nio shock (depending mainly onthe share of food in their CPI baskets), while global energy and non-fuel commodity pricesincrease. Therefore, we argue that macroeconomic policy formulation should take intoconsideration the likelihood and effects of El Nio weather episodes.

The rest of the paper is organized as follows. Section II. gives a brief description of theSouthern Oscillation cycle. Section III. describes the GVAR methodology and outlines ourmodelling approach. Section IV. investigates the macroeconomic effects of El Nio shocks.Finally, Section V. concludes and offers some policy recommendations.

II. THE SOUTHERN OSCILLATION

During "normal" years, a surface high pressure system develops over the coast of Peru and alow pressure system builds up in northern Australia and Indonesia (see Figure 2a). As a result,trade winds move strongly from east to west over the Pacific Ocean. These trade winds carrywarm surface waters westward and bring precipitation to Indonesia and Australia. Along thecoast of Peru, cold nutrient-rich water wells up to the surface, and thereby boosts the fishingindustry in South America.

However, in an El Nio year, air pressure drops along the coast of South America and overlarge areas of the central Pacific. The "normal" low pressure system in the western Pacificalso becomes a weak high pressure system, causing the trade winds to be reduced and

able to account for spillovers and the effects of ubserved and unobserved common factors (e.g. commodity-priceshocks and global finacial cycle)see Section III.A. for additional details.

4Changnon (1999) also argues that an El Nio event can benefit the economy of the United States on a netbasisamounting to 0.2% of GDP during the 1997/98 period.

7Figure 2. Southern Oscillation

(a) Southern Oscillation (b) El Nio Conditions

Source: Pidwirny (2006).

Figure 3. Global Climatological Effects of El Nino

Source: National Atmospheric and Oceanic Administrations (NOAA) Climate Prediction Center.

Figure 4. Southern Oscillation Index (Anomalies), 1979M42014M12

(a) SOI (b) SOI Anomalies

Source: Authors construction based on data from the Australian Bureau of Meteorology and the U.S. NationalOceanic and Atmospheric Administrations National Climatic Data Centre.Notes: Dashed-lines indicate thresholds for identifying El Nio and La Nia events.

8allowing the equatorial counter current (which flows west to east) to accumulate warm oceanwater along the coastlines of Peru (Figure 2b). This phenomenon causes the thermocline (theseparation zone between the mixed-layer above, much influenced by atmospheric fluxes, andthe deep ocean) to drop in the eastern part of Pacific Ocean, cutting off the upwelling of colddeep ocean water along the coast of Peru. Overall, the development of an El Nio bringsdrought to the western Pacific (including Australia), rains to the equatorial coast of SouthAmerica, and convective storms and hurricanes to the central Pacific. The globalclimatological effects of El Nio are summarized in Figure 3, showing the effects across twodifferent seasons. These changes in weather patterns have significant effects on agriculture,fishing, and construction industries, as well as on national and global commodity prices.Moreover, due to linkages of the Southern Oscillation with other climatic oscillations aroundthe world, El Nio effects reach far beyond the realm of the Pacific Ocean region.5

One of the ways of measuring El Nio intensity is by using the Southern Oscillation index(SOI), which is calculated based on air-pressure differentials in the South Pacific (betweenTahiti and Darwin). Sustained negative SOI values below -8 indicate El Nio episodes, whichtypically occur at intervals of three to seven years and last about two years. Figure 4 showsthat the 198283 and 199798 El Nios were quite severe (and had large adversemacroeconomic effects in many regions of the world), whereas other El Nios in our sampleperiod were relatively moderate: 1986-88, 1991-92, 1993, 1994-95, 2002-03, 2006-07, and2009-10. SOI "anomalies", which we use in our model, are defined as the deviation of the SOIindex from their historical averages and divided by their historical standard deviations.Sustained negative SOI anomaly values below -1 indicate El Nio episodes (Figure 4b).

III. MODELLING THE CLIMATE-MACROECONOMY RELATIONSHIP IN A GLOBALCONTEXT

We employ the Global VAR (GVAR) methodology to analyze the internationalmacroeconomic transmission of El Nio shocks. This framework takes into account both thetemporal and cross-sectional dimensions of the data; real and financial drivers of economicactivity; interlinkages and spillovers that exist between different regions; and the effects ofunobserved or observed common factors (e.g. energy and non-fuel commodity prices). This iscrucial as the impact of El Nio shocks cannot be reduced to one country but rather involve

5La Nia weather events (cold phases of the Southern Oscillation cycle) produce the opposite climate varia-tions from El Nio occurances. Since the effects of La Nia on fisheries along the coast of South America, whereEl Nio was named, are benign, they received relatively little attention.

9multiple regions, and may be amplified or dampened depending on the degree of openness ofthe countries and their trade structure. Before describing the data and our model specification,we provide a short exposition of the GVAR methodology below.

A. The Global VAR (GVAR) Methodology

We consider N + 1 countries in the global economy, indexed by i = 0; 1; :::; N . With theexception of the United States, which we label as 0 and take to be the reference country; allother N countries are modelled as small open economies. This set of individual VARX*models is used to build the GVAR framework. Following Pesaran (2004) and Dees et al.(2007), a VARX* (pi; qi) model for the ith country relates a ki 1 vector of domesticmacroeconomic variables (treated as endogenous), xit, to a ki 1 vector of country-specificforeign variables (taken to be weakly exogenous), xit:

i (L; pi)xit = ai0 + ai1t+i (L; qi)xit + uit; (1)

for t = 1; 2; :::; T , where ai0 and ai1 are ki 1 vectors of fixed intercepts and coefficients onthe deterministic time trends, respectively, and uit is a ki 1 vector of country-specificshocks, which we assume are serially uncorrelated with zero mean and a non-singularcovariance matrix, ii, namely uit s i:i:d: (0;ii). For algebraic simplicity, we abstract fromobserved global factors in the country-specific VARX* models. Furthermore,i (L; pi) = I

Ppii=1 iL

i and i (L; qi) =Pqi

i=0 iLi are the matrix lag polynomial of the

coefficients associated with the domestic and foreign variables, respectively. As the lag ordersfor these variables, pi and qi; are selected on a country-by-country basis, we are explicitlyallowing for i (L; pi) and i (L; qi) to differ across countries.

The country-specific foreign variables are constructed as cross-sectional averages of thedomestic variables using data on bilateral trade as the weights, wij:

xit =NXj=0

wijxjt; (2)

where j = 0; 1; :::N; wii = 0; andPN

j=0wij = 1. For empirical application, the trade weights

10

are computed as three-year averages:6

wij =Tij;2009 + Tij;2010 + Tij;2011Ti;2009 + Ti;2010 + Ti;2011

; (3)

where Tijt is the bilateral trade of country i with country j during a given year t and iscalculated as the average of exports and imports of country i with j, and Tit =

PNj=0 Tijt (the

total trade of country i) for t = 2009; 2010 and 2011; in the case of all countries.

Although estimation is done on a country-by-country basis, the GVAR model is solved for theworld as a whole, taking account of the fact that all variables are endogenous to the system asa whole. After estimating each country VARX*(pi; qi) model separately, all the k =

PNi=0 ki

endogenous variables, collected in the k 1 vector xt = (x00t;x01t; :::;x0Nt)0, need to be solvedsimultaneously using the link matrix defined in terms of the country-specific weights. To seethis, we can write the VARX* model in equation (1) more compactly as:

Ai (L; pi; qi) zit = 'it; (4)

for i = 0; 1; :::; N; where

Ai (L; pi; qi) = [i (L; pi)i (L; qi)] ; zit = (x0it;x0it)0 ;'it = ai0 + ai1t+ uit: (5)

Note that given equation (2) we can write:

zit = Wixt; (6)

where Wi = (Wi0;Wi1; :::;WiN), with Wii = 0, is the (ki + ki ) k weight matrix forcountry i defined by the country-specific weights, wij . Using (6) we can write (4) as:

Ai (L; p)Wixt = 'it; (7)

where Ai (L; p) is constructed from Ai (L; pi; qi) by settingp = max (p0; p1; :::; pN ; q0; q1; :::; qN) and augmenting the p pi or p qi additional terms inthe power of the lag operator by zeros. Stacking equation (7), we obtain the Global VAR(p)

6The main justification for using bilateral trade weights, as opposed to financial weights, is that the formerhave been shown to be the most important determinant of national business cycle comovements (see Baxter andKouparitsas (2005)).

11

model in domestic variables only:

G (L; p)xt = 't; (8)

where

G (L; p) =

0BBBBBBBBB@

A0 (L; p)W0

A1 (L; p)W1

.

.

.

AN (L; p)WN

1CCCCCCCCCA; 't =

0BBBBBBBBB@

'0t

'1t

.

.

.

'Nt

1CCCCCCCCCA: (9)

For an early illustration of the solution of the GVAR model, using a VARX*(1; 1) model, seePesaran (2004), and for an extensive survey of the latest developments in GVAR modeling,both the theoretical foundations of the approach and its numerous empirical applications, seeChudik and Pesaran (2014). The GVAR(p) model in equation (8) can be solved recursivelyand used for a number of purposes, such as forecasting or impulse response analysis.

Chudik and Pesaran (2013) extend the GVAR methodology to a case in which commonvariables are added to the conditional country models (either as observed global factors or asdominant variables). In such circumstances, equation (1) should be augmented by a vector ofdominant variables, !t, and its lag values:

i (L; pi)xit = ai0 + ai1t+i (L; qi)xit +i (L; si)!t + uit; (10)

where i (L; si) =Psi

i=0 iLi is the matrix lag polynomial of the coefficients associated

with the common variables. Here, !t can be treated (and tested) as weakly exogenous for thepurpose of estimation. The marginal model for the dominant variables can be estimated withor without feedback effects from xt: To allow for feedback effects from the variables in theGVAR model to the dominant variables via cross-section averages, we define the followingmodel for !t:

!t =

pwXl=1

!l!i;tl +pwXl=1

!lxi;tl + !t (11)

It should be noted that contemporaneous values of star variables do not feature in equation(11) and !t are "causal". Conditional (10) and marginal models (11) can be combined and

12

solved as a complete GVAR model as explained earlier.

B. Model Specification

Key countries in our sample include those likely to be directly affected by El Niomainlycountries in the Asia and Pacific region as well as those in the Americas, see Table 1 andSection II.. To investigate the possible indirect effects of El Nio (through trade, commodityprice and financial channels), we also include other major economies, such as Europeancountries, in the model. However, the main focus of the present study is not Europe, giventhat it is not likely to be directly affected by an El Nio shock. Therefore, for this empiricalapplication, we create a region consisting of 13 European countries. The time series data forthe Europe block are constructed as cross-sectionally weighted averages of the domesticvariables, using Purchasing Power Parity GDP weights, averaged over the 2009-2011 period.Thus, as displayed in Table 1, our model includes 33 countries (with 21country/region-specific models) covering over 90% of world GDP.

Table 1. Countries and Regions in the GVAR Model

Asia and Pacific North America EuropeAustralia Canada AustriaChina Mexico BelgiumIndia United States FinlandIndonesia FranceJapan South America GermanyKorea Argentina ItalyMalaysia Brazil NetherlandsNew Zealand Chile NorwayPhilippines Peru SpainSingapore SwedenThailand Middle East and Africa Switzerland

Saudi Arabia TurkeySouth Africa United Kingdom

We specify two different sets of individual country-specific models. The first model iscommon across all countries, apart from the United States. These 20 VARX* models includea maximum of six domestic variables (depending on whether data on a particular variable isavailable), or using the same terminology as in equation (1):

xit =yit; it; eqit; r

Sit; r

Lit; epit

0; (12)

13

where yit is the log of the real Gross Domestic Product at time t for country i, it is inflation,eqit is the log of real equity prices, rSit (rLit) is the short (long) term interest rate, and epit is thereal exchange rate. In addition, all domestic variables, except for that of the real exchangerate, have corresponding foreign variables computed as in equation (2):

xit =yit;

it; eq

it; r

Sit ; r

Lit

0: (13)

Following the GVAR literature, the twenty-first model (United States) is specified differently,mainly because of the dominance of the United States in the world economy. First, given theimportance of U.S. financial variables in the global economy, the U.S.-specific foreignfinancial variables, eqUS;t, rSUS;t, and rLUS;t, are not included in this model. The appropriatenessof exclusion of these variables was also confirmed by statistical tests, in which the weakexogeneity assumption was rejected for eqUS;t, rSUS;t, and rLUS;t. Second, since eit is expressedas the domestic currency price of a United States dollar, it is by construction determinedoutside this model. Thus, instead of the real exchange rate, we included eUS;t pUS;t as aweakly exogenous foreign variable in the U.S. model.7

Given our interest in analyzing the macroeconomic effects of El Nio shocks, we need toinclude the Southern Oscillation index anomalies (SOIt) in our framework. We model SOItas a dominant variable because there is no reason to believe that any of the macroeconomicvariables described above influences it. In other words, SOIt is included as a weaklyexogenous variable in each of the 21 country/region-specific VARX* models, with nofeedback effects from any of the macro variables to SOIt (hence a unidirectional causality).

Moreover, there is some anecdotal evidence that SOIt influences global commoditymarketsfor example, hot and dry summers in southeast Australia increases the frequencyand severity of bush fires, which reduces Australias wheat exports and thereby drives upglobal wheat prices, see Bennetton et al. (1998). We test this hypothesis formally byincluding the price of various commodities in our model. A key question is how should thesecommodity prices be included in the GVAR model? The standard approach to modellingcommodity markets in the GVAR literature (see Cashin et al. 2014) is to include the log ofnominal oil prices in U.S. dollars as a "global variable" determined in the U.S. VARX* model;that is the price of oil is included in the U.S. model as an endogenous variable while it istreated as weakly exogenous in the model for all other countries.8 The main justification for

7Weak exogeneity test results for all countries and variables are available upon request.

8An exception is Mohaddes and Pesaran (2015) which explicitly models the oil market as a dominant unit in

14

this approach is that the U.S. is the worlds largest oil consumer and a demand-side driver ofthe price of oil. However, it seems more appropriate for oil prices to be determined in globalcommodity markets rather in the U.S. model alone, given that oil prices are also affected by,for instance, any disruptions to oil supply in the Middle East.

Furthermore, given that El Nio events potentially affect the global prices of food, beverages,metals and agricultural raw materials, we also need to include the prices of these non-fuelcommodities in our model. However, rather than including the individual prices of non-fuelcommodities (such as wheat, coffee, timber, and nickel) we use a measure of real non-fuelcommodity prices in logs, pnft , constructed by the International Monetary Fund, with theweight of each of the 38 non-fuel commodities included in the index being equal to averageworld export earnings.9 Therefore, our commodity market model includes both the real crudeoil price (poilt ) and the real non-fuel commodity price as endogenous variables, where theformer can be seen as a good proxy for fuel prices in general. In addition, to capture theeffects of global economic conditions on world commodity markets, we include seven weaklyexogenous variables in this model. More specifically, real GDP, the rate of inflation, short andlong-term interest rates, real equity prices, and the real exchange rate are included as weaklyexogenous variables (constructed using purchasing power parity GDP weights, averaged over2009-2011), as is the SOIt.

IV. EMPIRICAL RESULTS

We obtain data on xit for the 33 countries included in our sample (see Table 1) from theGVAR website: https://sites.google.com/site/gvarmodelling, see Smith and Galesi (2014) formore details. Oil price data is also from the GVAR website, while data on non-fuelcommodity prices are from the International Monetary Funds International FinancialStatistics. Finally, the Southern Oscillation index (SOI) anomalies data are from NationalOceanic and Atmospheric Administrations National Climatic Data Centre. We use quarterlyobservations over the period 1979Q22013Q1 to estimate the 21 country-specificVARX*(pi; qi) models. However, prior to estimation, we determine the lag orders of thedomestic and foreign variables, pi and qi. For this purpose, we use the Akaike InformationCriterion (AIC) applied to the underlying unrestricted VARX* models. Given data

the GVAR framework.

9See http://www.imf.org/external/np/res/commod/table2.pdf for the details on these commodities and theirweights.

15

constraints, we set the maximum lag orders to pmax = qmax = 2. The selected VARX* ordersare reported in Table 2. Moreover, the lag order selected for the univariate SOIt model is 1and for the commodity price model is (1; 2), both based on the AIC.

Table 2. Lag Orders of the Country-Specific VARX*(p,q) Models Together with the Numberof Cointegrating Relations (r)

VARX* Order Cointegrating VARX* Order CointegratingCountry pi qi relations (ri) Country pi qi relations (ri)

Argentina 2 2 1 Malaysia 1 1 2Australia 1 1 4 Mexico 1 2 2Brazil 2 2 1 New Zealand 2 2 2Canada 1 2 2 Peru 2 2 1China 2 1 1 Philippines 2 1 2Chile 2 2 1 South Africa 2 2 3Europe 2 2 3 Saudi Arabia 2 1 1India 2 2 3 Singapore 2 1 1Indonesia 2 1 3 Thailand 1 1 1Japan 2 2 3 USA 2 2 2Korea 2 1 2

Notes: pi and qi denote the lag order for the domestic and foreign variables respectively and are selected by theAkaike Information Criterion (AIC). The number of cointegrating relations (ri) are selected using the maximaleigenvalue test statistics based on the 95% simulated critical values computed by stochastic simulations and 1000replications for all countries except for Korea and Saudi Arabia, for which we reduced ri below those suggestedby the maximal eigenvalue statistic to ensure that the persistence profiles were well behaved.Source: Authors estimations.

Having established the lag order of the 21 VARX* models, we proceed to determine thenumber of long-run relations. Cointegration tests with the null hypothesis of no cointegration,one cointegrating relation, and so on are carried out using Johansens maximal eigenvalue andtrace statistics as developed in Pesaran et al. (2000) for models with weakly exogenous I (1)regressors, unrestricted intercepts and restricted trend coefficients. We choose the number ofcointegrating relations (ri) based on the maximal eigenvalue test statistics using the 95%simulated critical values computed by stochastic simulations and 1000 replications.

We then consider the effects of system-wide shocks on the exactly-identified cointegratingvectors using persistence profiles developed by Lee and Pesaran (1993) and Pesaran and Shin(1996). On impact the persistence profiles (PPs) are normalized to take the value of unity, butthe rate at which they tend to zero provides information on the speed with which equilibriumcorrection takes place in response to shocks. The PPs could initially over-shoot, thusexceeding unity, but must eventually tend to zero if the vector under consideration is indeedcointegrated. In our analysis of the PPs, we noticed that the speed of convergence was very

16

slow for Korea and for Saudi Arabia where the system-wide shocks never really died out, sowe reduced ri by one for each country resulting in well behaved PPs overall. The finalselection of the number of cointegrating relations are reported in Table 2.

A. The Effects of El Nio on Real Output

In general, identification of shocks in economics is not a straightforward task, however, in ourapplication it is clear that the El Nio shock, a negative unit shock (equal to one standarderror) to SOI anomalies, SOIt, is identified by construction (as !t are "causal"). Table 3reports the estimated median impulse responses of real GDP growth to an El Nio shock,where the median responses on impact as well as the cumulated effects after the first, second,third, and fourth quarters are reported. The results show that an El Nio event has astatistically significant effect on real GDP growth for most of the countries in oursamplethere are only four countries for which the median effects are not statisticallysignificant at two or one standard deviations.10

As noted earlier, El Nio causes hot and dry summers in southeast Australia (Figure 3);increases the frequency and severity of bush fires; reduces wheat export, and drives up globalwheat prices. Exports and global prices of other commodities (food and raw agriculturalmaterials) are also affected by drought in Australia, further reducing output growth (theprimary sector constitutes 10% of Australias GDP, Table 4). New Zealand often experiencesdrought in parts of the country that are normally dry and floods in other places, resulting inlower agricultural output (the El Nio of 1997/98 was particularly severe in terms of outputloss for New Zealand). Therefore, it is not surprising that we observe a statistically-significantdrop in GDP growth of 0:37% for Australia and 0:29% for New Zealand, three and onequarters after an El Nio shock, respectively.11

Moreover, El Nio conditions usually coincide with a period of weak monsoon and risingtemperatures in India (see Figure 3) which adversely affects Indias agricultural sector andincreases domestic food prices. This is confirmed by our econometric analysis where IndiasGDP growth falls by 0:15% after the first quarter. The negative effect of El Nio is rathermuted in India, due to a number of mitigating factors. One such factor is the declining share

10Note that significance (for a particular variable and country) does not have to be seen on impact as the effectsof El Nio in most regions are felt during one specific season and hence could happen in a particular quarter ratherthan all quarters.

11See Kamber et al. (2013) for an analysis of the economic effects of drought in New Zealand.

17

Table 3. The Effects of an El Nio Shock on Real GDP Growth (in percent)

Country Impact Cumulated Responses After1 Quarter 2 Quarters 3 Quarters 4 Quarters

Argentina -0.08 0.03 0.29 0.64 1.08Australia -0.03 -0.18 -0.30 -0.37 -0.41Brazil -0.06 0.04 0.20 0.42 0.68Canada 0.00 0.13 0.33 0.58 0.85China -0.01 0.03 0.16 0.36 0.56Chile -0.19 -0.10 0.16 0.42 0.70Europe 0.02 0.09 0.27 0.49 0.69India -0.03 -0.15 -0.23 -0.25 -0.25Indonesia -0.35 -0.61 -0.91 -1.02 -1.01Japan -0.10 -0.12 0.01 0.20 0.37Korea 0.11 0.29 0.44 0.58 0.67Malaysia 0.08 0.06 0.13 0.27 0.43Mexico 0.03 0.37 0.71 1.12 1.57New Zealand -0.16 -0.29 -0.37 -0.42 -0.43Peru -0.07 -0.28 -0.35 -0.34 -0.33Philippines 0.06 0.09 0.11 0.17 0.21South Africa -0.11 -0.24 -0.47 -0.63 -0.72Saudi Arabia -0.09 -0.17 -0.14 0.00 0.18Singapore 0.09 0.28 0.54 0.87 1.18Thailand 0.47 0.78 1.11 1.49 1.81USA 0.05 0.10 0.23 0.39 0.55

Notes: Figures are median impulse responses to a one standard deviation reduction in SOI anomalies. The impactis in percentage points and the horizon is quarterly. Symbols ** and * denote significance at 595% and 1684%bootstrapped error bounds respectively.Source: Authors estimations.

Table 4. Share of Primary Sector in GDP (in percent), Averages over 2004-2013

Asia and Pacific North AmericaAustralia 10 Canada 10China 11 Mexico 12India 21 United States 3Indonesia 25Japan 1 South AmericaKorea 3 Argentina 11Malaysia 22 Brazil 7New Zealand 6 Chile 18Philippines 14 Peru 20Singapore 0Thailand 15 Africa

South Africa 10

Notes: Primary sector is the sum of agriculture, forestry, fishing and mining.Source: Haver.

18

of agricultural output in Indian GDP over timethe share of Indias primary sector in GDPwas 28% in 1997 and has dropped to 20% in 2013. The increase in the contribution of Rabicrops (sown in winter and harvested in the spring) and the decline in the contribution ofKharif crops (sown in the rainy monsoon season) over the past few decades is anothermitigating factor as sowing of Rabi crops is not directly affected by the monsoon.12 Notealso that the total irrigated area for major crops in India has increased from 22.6 millionhectares in 1950-51 to 86.4 million hectares in 2009-10. Moreover, due to more developedagricultural markets and policies, rising agriculture yield, and climatological early warningsystems, farmers are better able to switch to more drought-resistant and short-duration crops(with government assistance), at reasonably short notice. Furthermore, any severe rainfalldeficiency in India could have implications for public agricultural spending and governmentfinances. However, one should note that an El Nio year has not always resulted in weakmonsoons in India, see Saini and Gulati (2014).

Drought in Indonesia is also harmful for the local economy, and pushes up world prices forcoffee, cocoa, and palm oil, among other commodities. Furthermore, mining equipment inIndonesia relies heavily on hydropower; with deficient rain and low river currents, then lessnickel (which is used to strengthen steel) can be produced by the worlds top exporter ofnickel. Indonesian GDP growth falls by 0:91% at the end of the second quarter, and metalprices increase as global supply drops. This large growth effect is expected given that theshare of the primary sector (agricultural and mining) in Indonesian GDP is around 25 percent(see Table 4).

Looking beyond the Asia and Pacific region, South Africa also experiences hot and drysummers during an El Nio episode (Figure 3), which has adverse effects on its agriculturalproduction (the primary sector makes up 10% of South Africas GDP) with the empiricalresults suggesting a fall in GDP growth by 0:63% after the third quarter. Moreover, El Niotypically brings stormy winters in Chile and affects metal prices through supply chaindisruptionheavy rain in Chile will reduce access to its mountainous mining regions, wherelarge copper deposits are found. Therefore, we would expect an increase in metal prices and areduction in output growth, which we estimate to be 0:19% on impact. More frequenttyphoon strikes and cooler weather during summers are expected for Japan, which coulddepress consumer spending and growth. Our analysis suggests an initial drop of only 0:10%in Japanese output growth. However, we also observe that for both Chile and Japan, the

12In 1980-81 the ratio of Kharif to Rabi crop production was 1.5. In 2013-14 it is estimated at 0.95 (see, IndiaEconomic Survey 2014-15).

19

overall effect after four quarters is positive, by 0:70% and 0:37% respectively. This is mostlikely due to positive spillovers from their major trading partners. For instance, trade withChina, Europe, and the U.S. constitutes over 57% of each countrys total trade (see Table 5).The construction sector also sees a large boost following typhoons in Japan, which can partlyexplain the increase in growth after an initial decline. Finally, for northern Brazil, there is ahigh probability of a low rainfall year when El Nio is in force. Drought in northern parts ofBrazil can drive up world prices for coffee, sugar, and citrus. However, south-eastern Brazilgets plentiful rain in the spring/summer of an El Nio year, which leads to higher agriculturaloutput. We do not observe any significant effects for Brazil in the first two quarters,suggesting perhaps that the loss in agricultural output from drought in the northern part is tosome extent mitigated by above average yields in the south. More importantly, trade spilloversfrom other Latin American countries and systemic countries (China, Europe, and the U.S.)seem to suggest a positive overall effect on Brazil from an El Nio event in the third andfourth quarters following the shock.

El Nio years feature below-normal rainfall for the Philippines. However, the authorities haveextensive early-warning systems in place, including conservation management of the watersupply for Manila. As a result, we do not observe any significant growth effects. Moreover,the fisheries industry in Peru suffers because of the change in upwelling of nutrient-rich wateralong the coast. As Peru is the worlds largest exporter of fishmeal used in animal feed, alower supply from Peru has ramifications for livestock prices worldwide. However, at thesame time agricultural output in Peru rises due to the wetter weather. Although the mediangrowth effect for Peru is negative (0:33% after four quarters), it is in fact statisticallyinsignificant, so the positive growth effect from agricultural output (being 5:8% of GDP)offsets the negative impact on the fisheries industry (constituting 0:6% of GDP).

While an El Nio event results in lower growth for some economies, others may actuallybenefit due to lower temperatures, more rain, and less natural disasters. For instance, plentifulrains can help boost soybeans production in Argentina, which exports 95% of the soybeans itproduces, and for which the primary sector is around 11% of GDP (Table 4). Canada enjoyswarmer weather in an El Nio year, and in particular a greater return from its fisheries. Inaddition, the increase in oil prices means larger oil revenues for Canada, which is the worldsfifth-largest oil producer (averaging 3,856 million barrels per day in 2012). For Mexico weobserve less hurricanes on the east coast and more hurricanes on the west coast, which bringsgenerally stability to the oil sector and boosts exports (oil revenue is around 8% of GDP inMexico). For the United States, El Nio typically brings wet weather to California (benefitingcrops such as limes, almonds and avocados), warmer winters in the Northeast, increased

20

Table 5. Trade Weights, Averages over 20092011

Arge

ntin

a

Aust

ralia

Braz

il

Cana

da

Chin

a

Chile

Euro

pe

Indi

a

Indo

nesia

Japa

n

Kore

a

Mal

aysia

Mex

ico

New

Zea

land

Peru

Philip

pine

s

Sout

h Af

rica

Saud

i Ara

bia

Sing

apor

e

Thai

land

USA

Argentina 0.00 0.00 0.11 0.00 0.01 0.05 0.01 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.01 0.00 0.00 0.00 0.00Australia 0.01 0.00 0.01 0.00 0.04 0.01 0.03 0.04 0.03 0.06 0.04 0.04 0.00 0.24 0.00 0.02 0.02 0.01 0.03 0.05 0.01Brazil 0.32 0.01 0.00 0.01 0.03 0.08 0.04 0.02 0.01 0.01 0.02 0.01 0.01 0.00 0.06 0.00 0.02 0.02 0.01 0.01 0.02Canada 0.02 0.01 0.02 0.00 0.02 0.02 0.04 0.01 0.01 0.02 0.01 0.01 0.03 0.02 0.07 0.01 0.01 0.01 0.01 0.01 0.20China 0.13 0.25 0.19 0.08 0.00 0.24 0.25 0.16 0.14 0.27 0.28 0.16 0.09 0.16 0.19 0.12 0.18 0.15 0.14 0.16 0.18Chile 0.06 0.00 0.03 0.00 0.01 0.00 0.01 0.01 0.00 0.01 0.01 0.00 0.01 0.00 0.06 0.00 0.00 0.00 0.00 0.00 0.01Europe 0.21 0.15 0.28 0.12 0.23 0.19 0.00 0.30 0.11 0.14 0.12 0.13 0.08 0.16 0.20 0.13 0.38 0.19 0.14 0.15 0.22India 0.02 0.04 0.03 0.01 0.03 0.02 0.05 0.00 0.05 0.01 0.03 0.03 0.01 0.02 0.01 0.01 0.06 0.08 0.04 0.02 0.02Indonesia 0.01 0.03 0.01 0.00 0.02 0.00 0.01 0.04 0.00 0.04 0.04 0.05 0.00 0.02 0.00 0.03 0.01 0.02 0.10 0.05 0.01Japan 0.02 0.16 0.05 0.03 0.15 0.09 0.08 0.04 0.16 0.00 0.14 0.14 0.03 0.09 0.05 0.17 0.08 0.14 0.08 0.20 0.07Korea 0.02 0.07 0.04 0.01 0.10 0.06 0.04 0.04 0.08 0.08 0.00 0.05 0.03 0.04 0.04 0.07 0.03 0.11 0.07 0.04 0.03Malaysia 0.01 0.03 0.01 0.00 0.04 0.00 0.02 0.03 0.07 0.04 0.02 0.00 0.01 0.03 0.00 0.04 0.01 0.01 0.15 0.07 0.01Mexico 0.03 0.01 0.03 0.04 0.01 0.03 0.02 0.01 0.00 0.01 0.02 0.01 0.00 0.01 0.03 0.00 0.00 0.00 0.01 0.00 0.15New Zealand 0.00 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.00Peru 0.01 0.00 0.01 0.01 0.00 0.03 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00Philippines 0.01 0.00 0.00 0.00 0.01 0.00 0.01 0.00 0.01 0.02 0.01 0.02 0.00 0.01 0.00 0.00 0.00 0.01 0.03 0.02 0.01South Africa 0.01 0.01 0.01 0.00 0.01 0.00 0.03 0.03 0.01 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.01 0.01Saudi Arabia 0.00 0.01 0.02 0.00 0.02 0.00 0.03 0.07 0.02 0.04 0.05 0.01 0.00 0.02 0.00 0.03 0.04 0.00 0.03 0.03 0.02Singapore 0.00 0.05 0.01 0.00 0.03 0.00 0.03 0.05 0.14 0.03 0.04 0.15 0.00 0.04 0.00 0.12 0.01 0.04 0.00 0.06 0.02Thailand 0.01 0.04 0.01 0.00 0.03 0.01 0.02 0.02 0.05 0.05 0.02 0.07 0.01 0.03 0.01 0.06 0.02 0.03 0.05 0.00 0.01USA 0.10 0.09 0.16 0.67 0.19 0.17 0.27 0.13 0.09 0.17 0.13 0.12 0.68 0.12 0.23 0.16 0.11 0.16 0.11 0.11 0.00

Notes: Trade weights are computed as shares of exports and imports, displayed in columns by country (such thata column, but not a row, sum to 1).Source: International Monetary Funds Direction of Trade Statistics, 2009-2011.

21

rainfall in the South, diminished tornadic activity in the Midwest, and a decrease in thenumber of hurricanes that hit the East coast (see Figure 3). Therefore, not surprisingly, Table3 shows an increase in GDP growth of 1:08%, 0:85%, 1:57%, and 0:55% in the fourth quarterfollowing an El Nio shock for Argentina, Canada, Mexico, and the U.S., respectively. Theseestimates also take into account the positive spillover effects that an increase in U.S. GDPgrowth has on the Canadian and Mexican economies, given the extensive trade exposure ofthese two economies to the United States (trade weights are 67 and 68 percent respectively,see Table 5) as well as other third-market effects. The positive growth effect of 0:55% for theU.S. might seem large at first glance, however, it is not far from the estimated net benefits of$15 billion following the severe El Nio event of 1997-1998, which is equivalent to 0:2% ofGDP, see Changnon (1999). These net benefits are calculated based on a direct cost-benefitanalysis$4 billion (cost) and $19 billion (benefit)and a larger shock associated with the1997-98 El Nio event, but they do not take into account the indirect growth effects throughthird markets, which is captured in our GVAR framework.

Although El Nio is associated with dry weather in northern China and wet weather insouthern China (Figure 3), it is not clear that we should observe any direct positive or negativeeffects on Chinas output growth. In fact Table 3 shows that initially there are nostatistically-significant effects following an El Nio shock, but Chinese GDP growth increasesby 0:56% in the fourth quarter following an El Nio shock. This is mainly due to positivespillovers from trade with other major economiesChinese trade with the U.S. is about 19%of the total, and given that the U.S. is benefiting from an El Nio event, so does China.Moreover, a number of economies which are not directly affected by El Nio do benefit fromthe shock, mainly due to positive indirect spillovers from commercial trade and financialmarket links. For instance, Europe experiences an increase in real GDP growth of 0:69% inthe fourth quarter following an El Nio event, and Singapore by 1:18% (mainly due to anincrease in the shipping industry following the increase in demand from U.S. and other majoreconomies).

B. The Effects of El Nio on Real Commodity Prices

The higher temperatures and droughts following an El Nio event, particularly in Asia-Pacificcountries, not only increases the prices of non-fuel commodities (by 5:31% after four quarters,see Table 6), but also leads to higher demand for coal and crude oil as lower electricity outputis generated from both thermal power plants and hydroelectric dams. In addition, farmersincrease their water demand for irrigation purposes, which further increases the fuel demand

22

Table 6. The Effects of an El Nio Shock on Real Commodity Prices (in percent)

Series Impact Cumulated Responses After1 Quarter 2 Quarters 3 Quarters 4 Quarters

Non-Fuel Commodity Prices 0.42 0.77 1.97 3.75 5.31Oil Prices 1.20 4.23 7.80 11.09 13.87

Notes: Figures are median impulse responses to a one standard deviation reduction in SOI anomalies. The impactis in percentage points and the horizon is quarterly. Symbols ** and * denote significance at 595% and 1684%bootstrapped error bounds respectively.Source: Authors estimations.

for power generation and drives up energy prices. This is indeed confirmed here as crude oilprices (as a proxy for fuel prices) sustain a statistically significant and positive changefollowing an El Nio shock (see Table 6).

However, although the initial increase in oil prices arises from higher demand for power fromcountries such as India and Indonesia, oil prices remain high even four quarters after theinitial shock (Table 6). This is because an El Nio event has positive growth effects on majoreconomies (for example, China, European countries, and the U.S.) which demand more oil tobe able to sustain higher production. Therefore, what was initially an increase in oil prices dueto higher demand from Asia translates into a global oil demand shock (oil prices increasing atthe same time as global output rises; see Cashin et al. 2014 and Cashin et al. 2012 for details)a couple of quarters later. Excess demand also arises for non-fuel commodities (food,beverages, metals, and agricultural raw materials) and as a result their prices remainsignificant in the fourth quarter following an El Nio event, mainly due to lower supply fromthe Asia-Pacific region, but also due to higher global demand for non-fuel commodities.

C. The Effects of El Nio on Inflation

Turning to the inflationary effects of El Nio shocks, we find that for most countries in oursample, there exists statistically-significant upward pressure on inflation in the range of 0:09to 1:01 percentage points (Table 7). This is mainly due to higher fuel as well as non-fuelcommodity prices (Table 6), but is also the result of government policies (including bufferstock releases), inflation expectations, as well as aggregate demand-side pressures for thosecountries which experience a growth pick-up following an El Nio episode. Highest inflationjumps in Asia are observed in India (0:56% after three quarters), Indonesia (0:87% after twoquarters), and Thailand (0:55% after four quarters). These relatively large effects are due to

23

the high weight placed on food in the CPI basket of these countries: 47:6%, 32:7% and 33:5%,respectively. To examine this further we plot the weight of food in the CPI basket of the 20countries in our sample and the European region against the median impulse responses ofinflation to an El Nio shock in those countries. Figure 5 shows a clear positive relationshipbetween the two variables, with a correlation of 0.5, thereby providing further support to thenull hypothesis that inflation responses are larger in economies that have higher share of foodin their CPI baskets.

Note that production of perishables (i.e. fruits and vegetables) in India is affected less bymonsoon than food grains, while the prices of fruits and vegetables are relatively morevolatile. Moreover, inflation in food grains has historically been affected by governmentprocurement policies and administered minimum support prices in agriculture. During the lastdecade, inflation increased sharply after the 2009 drought in India, however, in the previousepisodes of drought in 2002 and 2004, inflation remained subdued. In 2009, droughtconditions were accompanied by a steep increase in minimum support prices, resulting in highfood grain inflation and consequently higher CPI inflation.13 Overall, government policies,tight monetary stances, high water reservoir levels, and excess food grain stocks could partlyoffset the inflationary impact of El Nio shocks on prices in India. For other Asianeconomies, which generally place lower weight on food in the CPI index, we notice a smallerincrease in inflation: China by 0:11% (32:5), Japan by 0:10% (24), Korea by 0:44% (13:9),Malaysia by 0:28% (30:3), and Philippines by 0:19% (39), with the numbers in bracketsrepresenting the weight of food as a percentage of total consumption in the CPI basket.

Inflation in the U.S. and Europe increases by smaller amounts 0:14 and 0:09 percentagepoints, respectively, but perhaps surprisingly Mexico sees an increase of 1:01% after twoquarters (with a 21 percent food share in its CPI basket). Finally, in South America inflationin the fourth quarter following an El Nio event increases by between 0:39% and 0:97%, but itis only statistically significant for Chile with an increase of 0:39%. There are only twocountries that experience a reduction in inflation following an El Nio eventNew Zealandby 0:61% after four quarters and Singapore by 0:07% on impact. For the former, this can beexplained by very large disinflation pressures during the initial occurrences of the El Nio(recessions, wage and price freezes, and structural reforms), and its well-anchored inflationexpectations14with an inflation target range of 13% on average over the medium-term and

13During the years 2002, 2004 and 2009 (all years of poor monsoons), CPI inflation averaged 4.1%, 3.9%, and12.3% in India, respectively.

14See also Buckle et al. (2002) for similar findings.

24

Table 7. The Effects of an El Nio Shock on Inflation (in percent)

Country Impact Cumulated Responses After1 Quarter 2 Quarters 3 Quarters 4 Quarters

Argentina 0.51 0.79 0.57 0.92 0.64Australia -0.01 0.02 0.02 0.01 0.00Brazil -0.30 -0.21 1.01 1.49 0.97Canada -0.05 -0.10 -0.08 -0.07 -0.07China 0.00 -0.02 0.00 0.06 0.11Chile 0.14 0.14 0.29 0.32 0.39Europe 0.00 0.00 0.02 0.06 0.09India 0.15 0.16 0.42 0.56 0.60Indonesia 0.25 0.61 0.87 0.95 0.91Japan 0.03 0.05 0.04 0.06 0.10Korea 0.01 0.12 0.22 0.34 0.44Malaysia 0.05 0.09 0.16 0.23 0.28Mexico 0.22 0.60 1.01 1.12 1.04New Zealand -0.06 -0.23 -0.39 -0.55 -0.61Peru -0.06 -0.73 -0.48 -0.38 0.65Philippines 0.11 0.06 0.19 0.22 0.27South Africa 0.10 -0.01 0.02 0.06 0.09Saudi Arabia 0.01 -0.03 -0.02 -0.01 -0.02Singapore -0.07 -0.06 -0.06 -0.06 -0.06Thailand 0.01 0.21 0.35 0.46 0.55USA 0.01 0.02 0.10 0.14 0.15

Notes: Figures are median impulse responses to a one standard deviation reduction in SOI anomalies. The impactis in percentage points and the horizon is quarterly. Symbols ** and * denote significance at 595% and 1684%bootstrapped error bounds respectively.Source: Authors estimations.

Figure 5. Food Weight in CPI Basket and Inflation Responses

Source: Authors calculations based on data from Haver and impulse response estimates in Table 7.

25

an average CPI inflation of around 2.5% since 1990.

D. Robustness Checks

To make sure that our results are not driven by the type of weights used to createcountry-specific foreign variable or solve the GVAR model as a whole, we experimentedusing Trade in Value Added (TiVA) weights (to account for supply chain factors) and foundthe impulse responses to be very similar to those with trade weights, wij , as used above.Therefore, as is now standard in the literature, we only report the results with the weightscalculated as the average of exports and imports of country i with j (Table 5). We alsoestimated our model with the foreign variables computed using trade weights averaged over2007-2009 and 2000-2013, and obtained very similar results to the benchmark weights(2009-2011) used in the earlier analysis. Moreover, we estimated a version of the modelsplitting the European region into Euro Area and 5 separate country VARX* models, therebyhaving a total of 26 country/region-specific VARX* models, and found the results to be robustto these changes. These results are not reported here, but are available on request.

V. CONCLUDING REMARKS

This paper contributed to the climate-macroeconomy literature by exploiting exogenousvariation in El Nio weather events over time to causatively identify the effects of El Nioshocks on growth, inflation, energy and non-fuel commodity prices. To analyze theinternational macroeconomic transmission of El Nio shocks we estimated a Global VAR(GVAR) model for 21 countries/regions over the period 1979Q22013Q1. Our modellingframework took into account real and financial drivers of economic activity; interlinkages andspillovers that exist between different regions; and the effects of unobserved or observedcommon factors (e.g. energy and non-fuel commodity prices). This is crucial as the impact ofEl Nio shocks cannot be reduced to one country, but rather involves multiple regions, andmay be amplified or reduced depending on the degree of openness of the countries and theirtrade structure.

We showed that there are considerable heterogeneities in the responses of different countriesto El Nio shocks. While Australia, Chile, Indonesia, India, Japan, New Zealand and SouthAfrica face a short-lived fall in economic activity following an El Nio weather shock, theUnited States, Europe and China actually benefit (possibly indirectly through third-market

26

effects) from such a climatological change. We also found that most countries in our sampleexperience short-run inflationary pressures following an El Nio episode, as global energyand non-fuel commodity prices increase.

The sensitivity of growth and inflation in different countries, as well as global commodityprices, to El Nio developments raises the question of which policies and institutions areneeded to counter the adverse effects of such shocks. These measures could include changesin the cropping pattern and input use (e.g. seeds of quicker-maturing crop varieties), rainwaterconservation, judicious release of food grain stocks, and changes in importspolicies/quantitiesthese measures would all help to bolster agricultural production inlow-rainfall El Nio years. On the macroeconomic policy side, any uptick in inflation arisingfrom El Nio shocks could be accompanied by a tightening of the monetary stance (ifsecond-round effects emerge), to help anchor inflation expectations. Investment in agriculturesector, mainly in irrigation, as well as building more efficient food value chains should also beconsidered in the longer-term. Our results also have policy implications for the design ofappropriate bands around inflation targets in countries that are directly affected by El Nioshocks. This depends on the share of food in their CPI basket and structural-food inflation, aswell as their susceptibility to El Nio shocks.

The research in this paper can be extended in a number of directions. A more complete modelfor the climate, including perhaps temperature, precipitation, storms, and other aspects of theweather, could be developed and integrated within our compact model of the world economy.This framework could then be utilized to investigate the effects of climate change and/orglobal weather shocks on economic activity. Modelling the global climate, however, is initself a major task and we shall therefore leave it as a task for future research.

27

REFERENCES

Adams, R. M., K. J. Bryant, B. A. Mccarl, D. M. Legler, J. OBrien, A. Solow, andR. Weiher (1995). Value of Improved Long-Range Weather Information. ContemporaryEconomic Policy 13(3), 1019.

Baxter, M. and M. A. Kouparitsas (2005). Determinants of Business Cycle Comovement: ARobust Analysis. Journal of Monetary Economics 52(1), pp. 113157.

Bennetton, J., P. Cashin, D. Jones, and J. Soligo (1998). An Economic Evaluation ofBushfire Prevention and Suppression. Australian Journal of Agricultural and ResourceEconomics 42(2), 149175.

Brunner, A. D. (2002). El Nino and World Primary Commodity Prices: Warm Water or HotAir? Review of Economics and Statistics 84(1), 176183.

Buckle, R. A., K. Kim, H. Kirkham, N. McLellan, and J. Sharma (2002). A Structural VARModel of the New Zealand Business Cycle. New Zealand Treasury Working Paper02/26.

Cashin, P., K. Mohaddes, and M. Raissi (2012). The Global Impact of the SystemicEconomies and MENA Business Cycles. IMF Working Paper WP/12/255.

Cashin, P., K. Mohaddes, M. Raissi, and M. Raissi (2014). The Differential Effects of OilDemand and Supply Shocks on the Global Economy. Energy Economics 44, 113134.

Changnon, S. A. (1999). Impacts of 1997-98 El Nio Generated Weather in the UnitedStates. Bulletin of the American Meteorological Society 80, 18191827.

Chudik, A. and M. H. Pesaran (2013). Econometric Analysis of High Dimensional VARsFeaturing a Dominant Unit. Econometric Reviews 32(5-6), 592649.

Chudik, A. and M. H. Pesaran (2014). Theory and Practice of GVAR Modeling. Journal ofEconomic Surveys, forthcoming.

Debelle, G. and G. Stevens (1995). Monetary Policy Goals for Inflation in Australia. ReserveBank of Australia Research Discussion Paper 9503.

Dees, S., F. di Mauro, M. H. Pesaran, and L. V. Smith (2007). Exploring the International

28

Linkages of the Euro Area: A Global VAR Analysis. Journal of AppliedEconometrics 22, 138.

Dell, M., B. F. Jones, and B. A. Olken (2014). What Do We Learn from the Weather? TheNew Climate-Economy Literature. Journal of Economic Literature 52(3), 74098.

Handler, P. and E. Handler (1983). Climatic Anomalies in the Tropical Pacific Ocean andCorn Yields in the United States. Science 220(4602), 11551156.

Hsiang, S. M., K. C. Meng, and M. A. Cane (2011). Civil Conflicts are Associated with theGlobal Climate. Nature 476, 438441.

Iizumi, T., J.-J. Luo, A. J. Challinor, G. Sakurai, M. Yokozawa, H. Sakuma, M. E. Brown,and T. Yamagata (2014). Impacts of El Nio Southern Oscillation on the Global Yieldsof Major Crops. Nature Communications 5.

Kamber, G., C. McDonald, and G. Price (2013). Drying Out: Investigating the EconomicEffects of Drought in New Zealand. Reserve Bank of New Zealand Analytical NoteSeries AN2013/02.

Lee, K. and M. H. Pesaran (1993). Persistence Profiles and Business Cycle Fluctuations in aDisaggregated Model of UK Output Growth. Ricerche Economiche 47, 293322.

Mohaddes, K. and M. H. Pesaran (2015). Oil Supply Shocks and the Global Economy: ACounterfactual Analysis. Cambridge Working Papers in Economics, forthcoming.

Pesaran, M. H. (2004). General Diagnostic Tests for Cross Section Dependence in Panels.IZA Discussion Paper No. 1240.

Pesaran, M. H. and Y. Shin (1996). Cointegration and Speed of Convergence to Equilibrium.Journal of Econometrics 71, 117143.

Pesaran, M. H., Y. Shin, and R. J. Smith (2000). Structural Analysis of Vector ErrorCorrection Models with Exogenous I(1) Variables. Journal of Econometrics 97,293343.

Pesaran, M. H., L. Vanessa Smith, and R. P. Smith (2007). What if the UK or Sweden hadJoined the Euro in 1999? An Empirical Evaluation Using a Global VAR. InternationalJournal of Finance & Economics 12(1), 5587.

29

Pidwirny, M. (2006). El Nio, La Nia and the Southern Oscillation. In Fundamentals ofPhysical Geography, 2nd Edition.

Saini, S. and A. Gulati (2014). El Nio and Indian Droughts - A Scoping Exercise. IndianCouncil for Research on International Economic Relations Working Paper 276.

Smith, L. V. and A. Galesi (2014). GVAR Toolbox 2.0. University of Cambridge: JudgeBusiness School.

Solow, A., R. Adams, K. Bryant, D. Legler, J. OBrien, B. McCarl, W. Nayda, and R. Weiher(1998). The Value of Improved ENSO Prediction to U.S. Agriculture. ClimaticChange 39(1), 4760.

Tol, R. S. J. (2009). The Economic Effects of Climate Change. Journal of EconomicPerspectives 23(2), 2951.

Ubilava, D. (2012). El Nio, La Nia, and World Coffee Price Dynamics. AgriculturalEconomics 43(1), 1726.

First Page and TOCGVAR_El_Nino_150331_IMFWPIntroductionThe Southern OscillationModelling the Climate-Macroeconomy Relationship in a Global Context The Global VAR (GVAR) MethodologyModel Specification

Empirical ResultsThe Effects of El Nio on Real OutputThe Effects of El Nio on Real Commodity PricesThe Effects of El Nio on InflationRobustness Checks

Concluding RemarksReferences